Abstract: Flynn’s taxonomy has significantly advanced our understanding and design of computer system architectures. However, its relevance to classifying computer systems in the era of artificial intelligence (AI) is limited. Currently, AI stands at a critical historical juncture: on one hand, it has achieved considerable progress; on the other hand, it faces challenges such as the unsustainable continuation of scaling laws. The future development path of AI remains highly uncertain. To address these issues, this article proposes constructing a taxonomy and evolutionary dynamics for computer systems from the perspective of computational utility and intelligence. The taxonomy aims to provide a systematic review of agents, while the evolutionary dynamics seeks to offer forward-looking insights into the future evolution of system distributions derived from this classification. By integrating both taxonomy and evolutionary dynamics, this approach aims to optimize and strategically plan the development pathway of AI in China.